【作者】 杨锦荣；

【导师】 张宪民；

【作者基本信息】 华南理工大学 ， 机械制造及其自动化， 2010， 硕士

【摘要】 当前电子产品朝着小型化、薄型化、批量小、品种多的方向发展,使得现代表面贴装技术(Surface Mounting Technology,SMT)的生产过程变得越来越复杂,传统产品质量管理渐渐变得难以适应日趋严格的质量管理要求。SMT生产线的数据采集系统中储存了海量的数据,这些数据背后隐含了大量的知识,而数据挖掘可以从大量的数据中智能地、自动地提取有价值的知识和信息。本论文针对现代SMT生产线的特点,从传统品质管理系统引入数据挖掘技术来研究和开发一种开放的、可扩展的、智能化的新型SMT生产的品质监控系统。本文引入数据挖掘技术的决策树算法,在分析SMT生产工艺的基础上建立了挖掘生产线数据的流程。然后,提出了一种离散化算法使决策树算法能处理连续值属性,并且引入了AdaBoost强化学习机思想来提高算法的精度。最后,获得改进的决策树算法可以有效地分析SMT生产线的历史数据,获取工艺参数评价的标准,辅助工作人员快速优化工艺参数。本文研究关联规则技术中的Apriori算法,引入有监督学习的思想,对Apriori算法进行改进,并应用于产品缺陷诊断。与原有Apriori算法相比,本文的算法效率更高,可以帮助工作人员快速解决质量缺陷问题,缩短解决质量异常的时间。最后,本文在Visual Basic.NET 2005集成开发环境,采用大型数据库系统SQL Server 2005结合与网络通讯技术对表面贴装技术产品质量监控系统进行了软件实现。并将系统应用到现场生产线上进行实验,实验结果表明了本系统的工艺参数评价功能有较高的准确度和缺陷诊断功能有良好的实用性。更多还原

【Abstract】 With electronic products’s development orientation toward smaller, thinner, small quantity and variety, Surface Mounting Technology (SMT) manufacturing process becomes more complex, traditional quality management becomes more and more difficult to adapt to the increasingly stringent quality management requirements. The SMT production line data acquisition system stored a mass amount of data, which implied a lot of knowledge; data mining can intelligently and automatically extract the valuable knowledge and information hidden in the large database. This thesis focuses on the specialty of modern SMT production line and the traditional production quality management system, using data mining to research and develop an open, scalable, intelligent new SMT production quality control system.This thesis introduced the decision tree algorithm of data mining technology, and based on the analysis of SMT production line to set up the flow of the mining production line data. Then, the discretization algorithm is proposed to let the decision tree algorithm in general be capable of handling continuous-valued attributes, also the AdaBoost reinforcement learning machine algorithm is introduced to improve the accuracy of the algorithm. Finally, we achieved an improved decision tree algorithm, which can effectively analyse the historical data of SMT production line to get the criteria evaluation of process parameters to support technicians for an optimization of the process parameters.In this thesis the author did a research about the application of Apriori algorithm, introduced the idea of supervised learning to improve the Apriori algorithm, and applied to diagnose the product defects. Finally, the improved algorithm we achieved is more efficient than the original Apriori algorithm, and can assist the technicians to faster solve the quality defects problem.Finally, in this thesis the development of the product quality control system software is base on the combination of Visual Basic.NET 2005 environment, SQL Server 2005 database system and network communication technology. After the experiment of the system on the real product line, the results indicated that the evaluation of process parameters function has a high accuracy and the defect diagnosis function has the right practicability.更多还原